问题
If multiple mappers are executed on the same node, will combiner combine the results from multiple mappers?
I can't find the answer for this in documents or books. And combiner examples I can find all seem to make a difference even if it can aggregate results from one mapper only.
回答1:
From Yahoo's Hadoop Tutorial:
The Combiner will receive as input all data emitted by the Mapper instances on a given node. The output from the Combiner is then sent to the Reducers, instead of the output from the Mappers. The Combiner is a "mini-reduce" process which operates only on data generated by one machine.
So to answer your question, yes.
回答2:
The wording in the Yahoo's tutorial (mentioned in Leonard's answer) seems to be misleading.
As of now, we get a combiner for every map task. The functionality suggested by this question is under development still: https://issues.apache.org/jira/browse/MAPREDUCE-4502
来源：https://stackoverflow.com/questions/21719308/does-combiner-work-on-results-from-multiple-mappers